Noise-Aware Machine Learning Accelerates Development of High-Latent-Heat Cu-Al-Ni Shape Memory Alloys for Thermal...

Area of Science:

• Materials Science
• Metallurgy
• Machine Learning Applications

Background:

• Copper-Aluminum-Nickel (Cu-Al-Ni) shape memory alloys (SMAs) are effective solid-solid phase-change materials (PCMs) for transient thermal management.
• Discovering high-latent-heat (ΔH) Cu-Al-Ni alloys is challenging due to the vast compositional space and noise in experimental data.

Purpose of the Study:

• To develop and apply a noise-aware machine learning strategy for accelerating the discovery of Cu-Al-Ni alloys with high ΔH and specific martensite start temperatures (Ms) between 100-200 °C.
• To improve the accuracy and reliability of data-driven screening using noisy experimental datasets.

Main Methods:

• Implemented a noise-aware machine learning strategy, specifically Kriging, to handle noisy experimental data.
• Optimized the noise level by minimizing prediction error to enhance model accuracy.
• Screened a wide compositional range of Cu-Al-Ni alloys to identify candidates with desired thermal properties.

Main Results:

• Discovered four new Cu-Al-Ni alloys with Ms ranging from 125 to 163 °C and ΔH from 9.27 to 9.86 J/g.
• The optimal alloy, Cu84Al13Ni3 (wt.%), exhibited Ms = 163 °C, ΔH = 9.86 J/g, thermal conductivity of 102 W·m⁻¹·K⁻¹, and a figure of merit (FOM) of 7272 × 10⁶ J² K⁻¹ s⁻¹ m⁻⁴.
• This alloy demonstrates an 11.8% increase in ΔH and a 33.75% increase in FOM compared to previous highest values in the target temperature range, showing excellent thermal cycling stability.

Conclusions:

• The noise-aware machine learning strategy effectively accelerates the discovery of high-performance Cu-Al-Ni SMAs for thermal management applications.
• The newly identified Cu-Al-Ni alloys offer superior thermal properties and stability, outperforming existing materials in the 100-200 °C range.
• This approach provides a reliable method for exploring complex material systems with noisy data.